Overview
For some time the Desktop Athena SPARC/Solaris configuration from Sun Microsystemshas been under review. It has reached the point where the costs of maintaining a second Athena desktop configuration are no longer justified by the incremental value of having Solaris as a second Athena desktop.
Over the summer of 2008 the Athena General Use Cluster number of Sun Solaris systems fell to 44. Through FY09 outreach will be made to customers to assist appropriately in dealing with the end of life of the Athena Desktop Solaris platform.
Status
- Goal: Begin last round of community communications by May 14, 2009
- End of Life 1 is targeted for July 1 2009
__
1) "End of Life" in this context means almost all public student lab/cluster Solaris Athena machines are pulled and replaced with Linux Athena systems; all centrally DLC purchased Solaris Athena machines are either handed over to the DLC for local support and transition to vendor Solaris, replaced with Linux Athena systems, or decommissioned
Areas of exposure
- Student use of Suns in Athena Clusters
Scale=small; Impact=low - Faculty who don't want to give up their Suns
Sscale=small; Impact=medium - Classes teaching with Sun-unique courseware
Scale=small; Impact=high - General use Solaris-only software:
- ArcInfo (GIS)
Out of date on Athena; available for personal Windows systems and on IS&T and DUSP WIN.MIT.EDU machines - FrameMaker
Will no longer be available; documents will need to be migrated to OpenOffice and Microsoft Office - Acrobat Distiller
Will no longer be available; PDFs can be generated using OpenOffice built-in PDF output configuration
- ArcInfo (GIS)
Additionally, we need to account for the impact on servers built from Athena Solaris. We need to identify the count of Solaris servers in the field and determine if migration or continuing to do security updates is the best approach. Note that the cost of Solaris support still drops even if only the desktop applications and services are discontinued initially, and if OS security updates are continued for a little while.
Solaris EOL Customer Options:
- Switch to Linux Athena hardware
- Migrate to work-alike software
- Go it alone and work directly with Sun for support
Plan of action
- Create Desktop Athena Solaris EOL Details web site telling users the timeline, rationale and what to do
- Create public Solaris Athena EOL letter for the remaining Solaris Athena community
- Directly deliver letter to known deliberate users
- Socialize EOL date for cluster Solaris Athena with OEIT, ACCORD, and other key stakeholders
- Establish MOTD messages to warn customers of Solaris Athena End of Life
- Display signs on machines in clusters warning of Solaris Athena End of Life
- Obtain list of faculty with Suns on their desks and get Hotline to begin migrating them ASAP. Let them forward problem customers to appropriate folks internally: othomas on point, alexp, amb, wdsouza for specific issues and resolutions
- Utilize mailing lists and beginning-of-term coordinated events to contact TAs and identify as many at-risk courses and course lockers as we can
- Make a list of the courses with Solaris-only exposure, and identify what to do
- Create a Solaris-sunset email list with members: othomas, jdreed, jmhunt, gettes, bowser and an OEIT representative to receive calls for help
- Quietly keep the 8 Solaris systems in 38-332 through the Fall term for classes that fall through the cracks
- Quietly keep Solaris dialups and x.dialup through Fall term for classes that fall through the cracks
2 Comments
Alex T Prengel
On the "go it alone" route- some users might ask if it's feasible to run OpenAFS on stock Solaris, and if locker software will run that way. Are we prepared to answer either part of the question? I don't especially want to be supporting old Sun software after we get rid of Athena Suns...
Oliver Thomas
It is difficult to say for sure without a handle on demand, but I suspect that we would not be able to sustain this model for long. End-of-Life for the platform also means EOL for 3rd-party locker software within a reasonable timeframe. Otherwise we will end up with an environment where we are trying to support software that we are unable to test on machines we do not have.
There are several short-term options available to prevent crises (mainly another semester of 37-332 and Solaris dialups) but beyond that moving to different software or locally installed and maintained systems with locally licensed and deployed software will need to be the solution if a specific DLC requires Solaris workstations.